Status: Model ready to be loaded into browser memory.Target: WebGPU
Load the model and enter an overwritten sentence below.
Runs completely locally on your GPU via WebGPU and Transformers.js.
Quick Presets:
Tip: Press Enter to send, Shift+Enter for new line
DeSlop Showcase(Multi-pass Pruning)
How it works: DeSlop is a 4-bit quantized Seq2Seq model running directly in your browser. It takes bloated, decorative, or AI-generated prose and non-destructively simplifies it through subtractive pruning. When passed repeatedly, sentences converge on concise, natural human prose.